Subject: | |
From: | |
Reply To: | |
Date: | Thu, 14 Feb 2008 14:50:01 -0600 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Title: Estimating Treatment Effects in the Presence of
Selection on Unobservables: The Generalized Endogenous Treatment
Model
Authors: Kevin Esterling, Michael Neblo, David Lazer
Entrydate: 2008-02-14 14:04:54
Keywords: Average Treatment Effects, Endogenous Treatment,
Selection on Unobservables, Latent Variables, Deliberation
Experiment, Political Efficacy
Abstract: In the presence of non-compliance, estimating
treatment effects in randomized experiments requires identifying
assumptions. Nonparametric matching estimators assume that
selection is on observables only. To address the case of
selection on unobservables, we propose the parametric
generalized endogenous treatment (GET) model to identify
treatment effects. The GET model uses participants' behavior
within an experiment to estimate the latent propensity to comply
with the treatment, and simultaneously estimates the correlation
between the treatment and a vector of substantive outcomes.
Using data from a recent deliberative field experiment, we
demonstrate how GET can be used to test hypotheses regarding
selection on unobservables; how matching remains biased when
selection is on unobservables; and that GET can retrieve similar
treatment effect estimates whether using matched data or the full
data set. We also show that the GET model can accommodate
non-response on follow up measures, multiple endogenous
regressors, and treatment effect heterogeneity.
http://polmeth.wustl.edu/retrieve.php?id=734
**********************************************************
Political Methodology E-Mail List
Editors: Melanie Goodrich, <[log in to unmask]>
Delia Bailey, <[log in to unmask]>
**********************************************************
Send messages to [log in to unmask]
To join the list, cancel your subscription, or modify
your subscription settings visit:
http://polmeth.wustl.edu/polmeth.php
**********************************************************
|
|
|